As a non-native English speaker myself, I don’t think that excuses using LLMs to write text for you. I’d much rather read broken English than Claudish. Fuck, just post straight in German or Spanish or Chinese, everybody has access to automated translation tools anyway.
That sounds reasonable to me, so we are talking about bad content after all, not simply it must be you without any resources to help you.
I completely acknowledge the personal and human aspect. But I still wonder, whether a technical discussion forum should be treated as if it were personal exchange between friends or a chess tournament?
My earlier point about resource mismatch and economic asymmetry has been lost, so I will repeat it.
If a company can afford lots of specialists (including communication specialists), who can also converse with each other, and these people each have an account on Julia Discourse, why should they be ok to make use of this “communication power” and say a solo pro or a small company should not be allowed to compensate for this using AI? Participation here can create visibility, reputation, package adoption, professional contacts, and commercial opportunities. I find it rather naive to implicitly assume that communication here takes places between academics solely in pursuit of science, which is how it looks to me.
It’s become a bit of a distration (read: point of extremely repetitive argument) that we have a rule about LLM output.
IMO the more important rule was always the “be concise” rule. I get the temptation to carve out the LLM sub-case: it’s now the main source of “textual littering” in producing extremely long meandering crap that nobody wants to engage with. It’s also less ambiguous to detect and flag.
However, we’ve actually had a problem with these sorts of long, unproductive posts since long before LLMs, and perhaps it was maybe even worse back then, because people used to actually read those long human generated posts, and often find something in the post to get angry at, and then spawn big kerfuffle. Nowadays, I think people are more and more tired of this style of text, and scroll past it more reflexively.
Maybe the solution is to just clamp down on long, meandering posts directly and leave the LLM aspect aside. But I get that this then gets into much more subjective territory that nobody wants to be responsible for deciding.
Even if you imagine that you are in some kind of competition with organizations in command of greater resources, posting slop doesn’t compensate for this condition; it makes you look worse and damages your reputation.
I agree with the other posters who said that it’s better to post imperfect English, or in your native language, than the vomit from some computer program. In fact, since this is supposed to be an international community, why take the knee to the hegemony of English? Write in whatever language you prefer. But they should be your words.
Oh I think you’ll find some people here are champing at the bit to decide those sorts of things. Personally, I’ve already had several messages censored in this thread.
Separately, I think perhaps a more pressing problem in this thread, and one that’s been a recurring problem in the community, is highly reactive, angry users who turn every conversation or comment into a trial about themselves, until it ends with some giant crash-out, where someone feels the need to fight a million other people, and everyone else feels some compulsion to weigh in and contribute to the dog-pile.
You’re the latest manifestation of this @kahliburke, but far from the first and likely not the last. It seems like our Discourse community attracts this kind of problem much more often than our ‘chat’ based communities. I think we as a community need to think about what is causing this dynamic to be so much worse here on Discourse, and what we can do about it.
Please don’t! This put the onus of translation onto (almost) every single reader. This is one of the most impolite things you can do here IMO.
Furthermore I have been moderated for writing two or three words (nothing bad) in Russian, so it is apparently not generally accepted.
I’ve been told that this topic has nothing to do with censorship.
I disagree, when the original post says, and I quote:
So, it is suggested that we use LLMs (which are allegedly “sloppy”) to infer whether text was produced using LLMs… and then delete it automatically. Is that not related to censoring? Am I just really confused here?
@kahliburke — this is a moderated forum. Period. Full stop. That means that moderators can and do take actions to try to keep threads on-topic. This will not change.
Complaining about the very existence of such moderation is always off-topic.
I do agree. As nice as it sounds in theory that everyone sticks to their most comfortable language, that leaves the misinterpretations to a machine. So in practice that is worse than me not being able to write English.
Write English, be a bit off in meaning here and there and grant the same generosity to others when reading their personal writings.
Question for you, was this thread originally placed on the “Off Topic” channel?
Second question, is mentioning the topic of censorship relevant in a topic that literally suggests automated deletion of user posts?
This topic is about our policies around moderation of LLM outputs. Yes, it was initially in Offtopic, and I moved it to Meta Discussion, but even in Off-topic we try to keep topics focused in a single point of discussion.
To wit: Let’s talk about automated translation in Should we enable automatic natural language translation here?.
Except that it is not well structured, using bullet points does not imply structured writing.
If you were to give me a snippet of your conceivably good AI output I am pretty sure I could rewrite an entire paragraph into one a single cogent sentence. The problem with AI prose is simply that it’s bad in that it has no regard for the audience. It somehow manages to be both incredibly dense and meandering at the same time. I should be able to judge how much information is contained in a post from the number of words, with AI prose this is simply not possible anymore. This all puts the onus on the reader to do extra processing that should have been done by the writer in the first place.